llvm
/
torch-mlir
mirror of https://github.com/llvm/torch-mlir
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Simplify matrix configuration for CI workflows (#1213) 

Addresses https://github.com/llvm/torch-mlir/issues/1207. 

#### Provisioned jobs:
```
# ubuntu - x86_64 - llvm in-tree     - pytorch binary - build+test    # most used dev flow and fastest signal
# ubuntu - x86_64 - llvm out-of-tree - pytorch source - build+test    # most elaborate build
# macos  - arm64  - llvm in-tree     - pytorch source - build only    # cross compile, can't test arm64
```

#### Main changes
- [x] Spawn macos builds from a separate matrix (in the same workflow). It made sense to do this as they are fairly different from ubuntu (cross compile, use a different cmake configuration). This simplifies the matrix configuration and exclusions quite a bit, and makes the workflow a bit more tractable and maintenance friendly.
- [x] Remove the submodule md5sum step for ccache config. This was [broken](https://github.com/llvm/torch-mlir/runs/7779288734?check_suite_focus=true#step:3:145) for a while now.
- [x] Removes unused matrix options - `os`, `targetarch`, `python-version`, `llvmtype`.
- [x] Address ZSTD [comment](https://github.com/llvm/torch-mlir/pull/1204#discussion_r942349282) on @powderluv's cross compile [PR](https://github.com/llvm/torch-mlir/pull/1204). 

#### Further improvements (to be addressed in follow-on):
* ubuntu-x86_64 out-of-tree integration tests fail ([error](https://github.com/sjain-stanford/torch-mlir/runs/7781264029?check_suite_focus=true)); only run unit tests for now (tests are excluded in current CI too)

#### Passing workflow:
https://github.com/sjain-stanford/torch-mlir/actions/runs/2840676309
![image](https://user-images.githubusercontent.com/19234106/184194535-f3807991-401a-4cb9-b030-0ee8c334eba3.png)
pull/1215/head snapshot-20220812.562
Sambhav Jain 2022-08-11 16:35:15 -07:00 committed by GitHub
parent 51bfe25c89
commit f00ca91db0
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
4 changed files with 133 additions and 119 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
.github/actions/setup-build/action.yml vendored
View File

@ -11,27 +11,28 @@ inputs:
runs: runs:
using: "composite" using: "composite"
steps: steps:
- name: Set up Python - name: Set up Python
uses: actions/setup-python@v2 uses: actions/setup-python@v2
with: with:
python-version: 3.9 python-version: 3.9
- name: Install MLIR Python depends - name: Install MLIR Python depends
run: | run: |
python -m pip install -r $GITHUB_WORKSPACE/externals/llvm-project/mlir/python/requirements.txt python -m pip install -r $GITHUB_WORKSPACE/externals/llvm-project/mlir/python/requirements.txt
shell: bash shell: bash
- name: Install PyTorch nightly depends - name: Install PyTorch nightly depends
run: | run: |
python -m pip install -r requirements.txt python -m pip install -r requirements.txt
shell: bash shell: bash
- name: Install Ninja - name: Install Ninja
uses: llvm/actions/install-ninja@55d844821959226fab4911f96f37071c1d4c3268 uses: llvm/actions/install-ninja@55d844821959226fab4911f96f37071c1d4c3268
- name: Get Submodule Hash
id: get-submodule-hash
run: echo "::set-output name=hash::$(md5sum $(git submodule status))"
shell: bash
- name: Ccache for C++ compilation - name: Ccache for C++ compilation
uses: hendrikmuhs/ccache-action@85bd285251b831e5a761d26e3dbfdbccbca1b23f uses: hendrikmuhs/ccache-action@v1.2
with: with:
key: ${{ runner.os }}-clangreleaseasserts-${{ steps.get-submodule-hash.outputs.hash }}${{ inputs.cache-suffix }} key: ${{ runner.os }}-torch_mlir_build_assets-${{ inputs.cache-suffix }}
max-size: 2G max-size: 2G

15
.github/workflows/bazelBuildAndTest.yml vendored
View File

@ -2,27 +2,30 @@ name: Bazel Build and Test
on: on:
push: push:
branches: branches: [ main ]
- main
workflow_dispatch: workflow_dispatch:
jobs: jobs:
build: ubuntu-build:
name: Build and Test (Release Asserts) name: ubuntu-x86_64
runs-on: ubuntu-20.04 runs-on: ubuntu-22.04
steps: steps:
- name: Set up Python - name: Set up Python
uses: actions/setup-python@v2 uses: actions/setup-python@v2
with: with:
python-version: 3.9 python-version: 3.9
- name: Get torch-mlir
- name: Checkout torch-mlir
uses: actions/checkout@v2 uses: actions/checkout@v2
with: with:
submodules: 'true' submodules: 'true'
- name: Build with bazel - name: Build with bazel
run: | run: |
cd $GITHUB_WORKSPACE/utils/bazel cd $GITHUB_WORKSPACE/utils/bazel
bazel build @torch-mlir//... bazel build @torch-mlir//...
- name: Send mail - name: Send mail
if: failure() if: failure()
uses: dawidd6/action-send-mail@v3 uses: dawidd6/action-send-mail@v3

223
.github/workflows/buildAndTest.yml vendored
View File

@ -1,182 +1,193 @@
name: Build and Test name: Build and Test
on: on:
push:
branches:
- main
pull_request: pull_request:
branches: [ main ]
push:
branches: [ main ]
workflow_dispatch: workflow_dispatch:
# Provisioned Jobs:
# ubuntu - x86_64 - llvm in-tree - pytorch binary - build+test # most used dev flow and fastest signal
# ubuntu - x86_64 - llvm out-of-tree - pytorch source - build+test # most elaborate build
# macos - arm64 - llvm in-tree - pytorch source - build only # cross compile, can't test arm64
jobs: jobs:
build-validate: ubuntu-build:
name: ubuntu-x86_64
runs-on: ubuntu-22.04
strategy: strategy:
fail-fast: true fail-fast: true
matrix: matrix:
os: [macos-12, ubuntu-22.04] llvm-build: [in-tree, out-of-tree]
targetarch: [x86_64, arm64]
python-version: ["3.10"]
torch-binary: [ON, OFF] torch-binary: [ON, OFF]
llvmtype: [source, binary]
llvmbuildtype: [in-tree, out-of-tree]
exclude: exclude:
# No need for "out-of-tree LLVM and PyTorch source" # Exclude llvm in-tree and pytorch source
- llvmtype: source - llvm-build: in-tree
llvmbuildtype: in-tree
- llvmtype: binary
llvmbuildtype: out-of-tree
- llvmbuildtype: out-of-tree
torch-binary: OFF torch-binary: OFF
# Disable M1 builds until https://github.com/llvm/torch-mlir/issues/1094 is fixed # Exclude llvm out-of-tree and pytorch binary
- targetarch: arm64 - llvm-build: out-of-tree
os: ubuntu-22.04 torch-binary: ON
# macOS we only do source builds to reduce options
- os: macos-12
torch-binary: OFF
- os: macos-12
llvmtype: source
- os: macos-12
llvmtype: out-of-tree
- os: macos-12
targetarch: x86_64
runs-on: ${{ matrix.os }}
steps: steps:
- name: Checkout torch-mlir - name: Checkout torch-mlir
uses: actions/checkout@v2 uses: actions/checkout@v2
with: with:
submodules: 'true' submodules: 'true'
- uses: ./.github/actions/setup-build
- name: Setup ccache
uses: ./.github/actions/setup-build
with: with:
cache-suffix: ${{ matrix.os }}-${{ matrix.targetarch }}-${{ matrix.llvmtype }}-${{ matrix.llvmbuildtype }} cache-suffix: ubuntu-x86_64-${{ matrix.llvm-build }}-${{ matrix.torch-binary }}
- name: Configure llvm-cross-compile - name: Configure llvm-build='in-tree' torch-binary='${{ matrix.torch-binary}}'
# libzstd on GH Runners are only x86_64 to remove them. # Fastest build, most used dev flow
if: matrix.targetarch == 'arm64' if: matrix.llvm-build == 'in-tree'
run: | run: |
sudo rm -rf /usr/local/lib/libzstd.*.dylib
sudo rm -rf /usr/local/lib/cmake/zstd/*
cd $GITHUB_WORKSPACE
cmake -GNinja -Bbuild_${{ matrix.targetarch }} \
-DCMAKE_BUILD_TYPE=Release \
-DCMAKE_LINKER=lld \
-DCMAKE_C_COMPILER_LAUNCHER=ccache -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ \
-DPython3_EXECUTABLE=$(which python) \
-DLLVM_ENABLE_ASSERTIONS=ON \
-DLLVM_ENABLE_PROJECTS=mlir \
-DLLVM_EXTERNAL_PROJECTS="torch-mlir;torch-mlir-dialects" \
-DLLVM_EXTERNAL_TORCH_MLIR_SOURCE_DIR="$GITHUB_WORKSPACE" \
-DLLVM_EXTERNAL_TORCH_MLIR_DIALECTS_SOURCE_DIR="${GITHUB_WORKSPACE}/external/llvm-external-projects/torch-mlir-dialects" \
-DMLIR_ENABLE_BINDINGS_PYTHON=ON \
-DTORCH_MLIR_ENABLE_LTC=OFF \
-DTORCH_MLIR_ENABLE_MHLO=OFF \
-DTORCH_MLIR_USE_INSTALLED_PYTORCH=${{ matrix.torch-binary }} \
-DCMAKE_OSX_ARCHITECTURES=${{ matrix.targetarch }} \
-DMACOSX_DEPLOYMENT_TARGET=12.0 \
-DLLVM_TARGETS_TO_BUILD="AArch64" \
-DLLVM_USE_HOST_TOOLS=ON \
$GITHUB_WORKSPACE/externals/llvm-project/llvm
- name: Configure llvm-binary-torch-src-or-binary
# Should be the fastest builds for CI and fails fast
# OSX CMake flags are ignored on Linux
if: matrix.llvmtype == 'binary'
run: |
cd $GITHUB_WORKSPACE
cmake -GNinja -Bbuild \ cmake -GNinja -Bbuild \
-DCMAKE_BUILD_TYPE=Release \ -DCMAKE_BUILD_TYPE=Release \
-DCMAKE_C_COMPILER=clang \
-DCMAKE_CXX_COMPILER=clang++ \
-DCMAKE_C_COMPILER_LAUNCHER=ccache \
-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DCMAKE_LINKER=lld \ -DCMAKE_LINKER=lld \
-DCMAKE_C_COMPILER_LAUNCHER=ccache -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ \
-DPython3_EXECUTABLE=$(which python) \
-DLLVM_ENABLE_ASSERTIONS=ON \ -DLLVM_ENABLE_ASSERTIONS=ON \
-DLLVM_ENABLE_PROJECTS=mlir \ -DLLVM_ENABLE_PROJECTS=mlir \
-DLLVM_EXTERNAL_PROJECTS="torch-mlir;torch-mlir-dialects" \ -DLLVM_EXTERNAL_PROJECTS="torch-mlir;torch-mlir-dialects" \
-DLLVM_EXTERNAL_TORCH_MLIR_SOURCE_DIR="$GITHUB_WORKSPACE" \ -DLLVM_EXTERNAL_TORCH_MLIR_SOURCE_DIR="$GITHUB_WORKSPACE" \
-DLLVM_EXTERNAL_TORCH_MLIR_DIALECTS_SOURCE_DIR="${GITHUB_WORKSPACE}/external/llvm-external-projects/torch-mlir-dialects" \ -DLLVM_EXTERNAL_TORCH_MLIR_DIALECTS_SOURCE_DIR="${GITHUB_WORKSPACE}/external/llvm-external-projects/torch-mlir-dialects" \
-DLLVM_TARGETS_TO_BUILD=host \
-DMLIR_ENABLE_BINDINGS_PYTHON=ON \ -DMLIR_ENABLE_BINDINGS_PYTHON=ON \
-DTORCH_MLIR_ENABLE_MHLO=ON \ -DTORCH_MLIR_ENABLE_MHLO=ON \
-DTORCH_MLIR_USE_INSTALLED_PYTORCH=${{ matrix.torch-binary }} \ -DTORCH_MLIR_USE_INSTALLED_PYTORCH="${{ matrix.torch-binary }}" \
-DCMAKE_OSX_ARCHITECTURES=${{ matrix.taregetarch }} \ -DPython3_EXECUTABLE="$(which python)" \
-DMACOSX_DEPLOYMENT_TARGET=10.15 \
-DLLVM_TARGETS_TO_BUILD=host \
$GITHUB_WORKSPACE/externals/llvm-project/llvm $GITHUB_WORKSPACE/externals/llvm-project/llvm
- name: Configure llvm-source-out-of-tree-torch-src-or-binary - name: Configure llvm-build='out-of-tree' torch-binary='${{ matrix.torch-binary}}'
# This build takes a while but is expected to almost always be cached. # Most elaborate build, but cached
# A cache invalidation occurs when the committed LLVM version is changed. # A cache invalidation occurs when the committed LLVM version is changed.
if: matrix.llvmtype == 'source' if: matrix.llvm-build == 'out-of-tree'
run: | run: |
cd $GITHUB_WORKSPACE
cmake -GNinja -Bllvm-build \ cmake -GNinja -Bllvm-build \
-DCMAKE_BUILD_TYPE=Release \ -DCMAKE_BUILD_TYPE=Release \
-DCMAKE_C_COMPILER=clang \
-DCMAKE_CXX_COMPILER=clang++ \
-DCMAKE_C_COMPILER_LAUNCHER=ccache \
-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DCMAKE_LINKER=lld \ -DCMAKE_LINKER=lld \
-DCMAKE_C_COMPILER_LAUNCHER=ccache -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ \
-DPython3_EXECUTABLE=$(which python) \
-DLLVM_ENABLE_ASSERTIONS=ON \ -DLLVM_ENABLE_ASSERTIONS=ON \
-DLLVM_ENABLE_PROJECTS=mlir \ -DLLVM_ENABLE_PROJECTS=mlir \
-DMLIR_ENABLE_BINDINGS_PYTHON=ON \
-DLLVM_TARGETS_TO_BUILD=host \ -DLLVM_TARGETS_TO_BUILD=host \
externals/llvm-project/llvm -DMLIR_ENABLE_BINDINGS_PYTHON=ON \
-DPython3_EXECUTABLE="$(which python)" \
$GITHUB_WORKSPACE/externals/llvm-project/llvm
cmake --build llvm-build cmake --build llvm-build
# TODO: Reenable LTC once OOT build is successful (https://github.com/llvm/torch-mlir/issues/1154) # TODO: Reenable LTC once OOT build is successful (https://github.com/llvm/torch-mlir/issues/1154)
cmake -GNinja -Bbuild \ cmake -GNinja -Bbuild \
-DCMAKE_C_COMPILER=clang \
-DCMAKE_CXX_COMPILER=clang++ \
-DCMAKE_C_COMPILER_LAUNCHER=ccache \
-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DCMAKE_LINKER=lld \ -DCMAKE_LINKER=lld \
-DCMAKE_C_COMPILER_LAUNCHER=ccache -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \ -DLLVM_TARGETS_TO_BUILD=host \
-DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ \ -DLLVM_DIR="$GITHUB_WORKSPACE/llvm-build/lib/cmake/llvm/" \
-DMLIR_DIR="$(pwd)/llvm-build/lib/cmake/mlir/" \ -DMLIR_DIR="$GITHUB_WORKSPACE/llvm-build/lib/cmake/mlir/" \
-DLLVM_DIR="$(pwd)/llvm-build/lib/cmake/llvm/" \
-DMLIR_ENABLE_BINDINGS_PYTHON=OFF \ -DMLIR_ENABLE_BINDINGS_PYTHON=OFF \
-DTORCH_MLIR_ENABLE_MHLO=ON \ -DTORCH_MLIR_ENABLE_MHLO=ON \
-DTORCH_MLIR_USE_INSTALLED_PYTORCH=${{ matrix.torch-binary }} \ -DTORCH_MLIR_USE_INSTALLED_PYTORCH=${{ matrix.torch-binary }} \
-DTORCH_MLIR_ENABLE_LTC=OFF \ -DTORCH_MLIR_ENABLE_LTC=OFF \
-DPython3_EXECUTABLE=$(which python) \ -DPython3_EXECUTABLE="$(which python)" \
-DLLVM_TARGETS_TO_BUILD=host \ $GITHUB_WORKSPACE
.
- name: Build torch-mlir - name: Build torch-mlir
if: matrix.targetarch == 'x86_64'
run: | run: |
cmake --build build cmake --build build
- name: Build torch-mlir (cross-compile)
if: matrix.targetarch == 'arm64'
run: |
cmake --build build_${{ matrix.targetarch }}
- name: Run torch-mlir unit tests - name: Run torch-mlir unit tests
if: matrix.llvmtype == 'binary'
run: | run: |
cd $GITHUB_WORKSPACE
export PYTHONPATH="$GITHUB_WORKSPACE/build/tools/torch-mlir/python_packages/torch_mlir" export PYTHONPATH="$GITHUB_WORKSPACE/build/tools/torch-mlir/python_packages/torch_mlir"
cmake --build build --target check-torch-mlir-all cmake --build build --target check-torch-mlir-all
- name: Run RefBackend - TorchScript end-to-end tests - name: Run refbackend e2e integration tests
if: matrix.llvmtype == 'binary' if: matrix.llvm-build == 'in-tree'
run: | run: |
cd $GITHUB_WORKSPACE
export PYTHONPATH="$GITHUB_WORKSPACE/build/tools/torch-mlir/python_packages/torch_mlir" export PYTHONPATH="$GITHUB_WORKSPACE/build/tools/torch-mlir/python_packages/torch_mlir"
python -m e2e_testing.torchscript.main --config=refbackend -v python -m e2e_testing.torchscript.main --config=refbackend -v
- name: Run EagerMode - TorchScript end-to-end tests - name: Run eager_mode e2e integration tests
if: matrix.llvmtype == 'binary' if: matrix.llvm-build == 'in-tree'
run: | run: |
cd $GITHUB_WORKSPACE
export PYTHONPATH="$GITHUB_WORKSPACE/build/tools/torch-mlir/python_packages/torch_mlir" export PYTHONPATH="$GITHUB_WORKSPACE/build/tools/torch-mlir/python_packages/torch_mlir"
python -m e2e_testing.torchscript.main --config=eager_mode -v python -m e2e_testing.torchscript.main --config=eager_mode -v
- name: Run TOSA backend - TorchScript end-to-end tests - name: Run tosa e2e integration tests
if: matrix.llvmtype == 'binary' if: matrix.llvm-build == 'in-tree'
run: | run: |
cd $GITHUB_WORKSPACE
export PYTHONPATH="$GITHUB_WORKSPACE/build/tools/torch-mlir/python_packages/torch_mlir" export PYTHONPATH="$GITHUB_WORKSPACE/build/tools/torch-mlir/python_packages/torch_mlir"
python -m e2e_testing.torchscript.main --config=tosa -v python -m e2e_testing.torchscript.main --config=tosa -v
- name: Run Lazy Tensor Core - TorchScript end-to-end tests - name: Run lazy_tensor_core e2e integration tests
if: matrix.llvmtype == 'binary' if: matrix.llvm-build == 'in-tree'
run: | run: |
cd $GITHUB_WORKSPACE
export PYTHONPATH="$GITHUB_WORKSPACE/build/tools/torch-mlir/python_packages/torch_mlir" export PYTHONPATH="$GITHUB_WORKSPACE/build/tools/torch-mlir/python_packages/torch_mlir"
python -m e2e_testing.torchscript.main --config=lazy_tensor_core -v python -m e2e_testing.torchscript.main --config=lazy_tensor_core -v
macos-build:
name: macos-arm64
runs-on: macos-12
strategy:
fail-fast: true
matrix:
llvm-build: [in-tree, out-of-tree]
torch-binary: [ON, OFF]
exclude:
# Exclude llvm in-tree and pytorch binary
- llvm-build: in-tree
torch-binary: ON
# Exclude llvm out-of-tree altogether
- llvm-build: out-of-tree
steps:
- name: Checkout torch-mlir
uses: actions/checkout@v2
with:
submodules: 'true'
- name: Setup ccache
uses: ./.github/actions/setup-build
with:
cache-suffix: macos-arm64-${{ matrix.llvm-build }}-${{ matrix.torch-binary }}
- name: Configure llvm-build='in-tree' torch-binary='${{ matrix.torch-binary}}'
# libzstd on GH Runners are only x86_64 to remove them.
run: |
cmake -GNinja -Bbuild_arm64 \
-DCMAKE_BUILD_TYPE=Release \
-DCMAKE_C_COMPILER=clang \
-DCMAKE_CXX_COMPILER=clang++ \
-DCMAKE_C_COMPILER_LAUNCHER=ccache \
-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DCMAKE_LINKER=lld \
-DCMAKE_OSX_ARCHITECTURES=arm64 \
-DLLVM_ENABLE_ASSERTIONS=ON \
-DLLVM_ENABLE_PROJECTS=mlir \
-DLLVM_EXTERNAL_PROJECTS="torch-mlir;torch-mlir-dialects" \
-DLLVM_EXTERNAL_TORCH_MLIR_SOURCE_DIR="$GITHUB_WORKSPACE" \
-DLLVM_EXTERNAL_TORCH_MLIR_DIALECTS_SOURCE_DIR="${GITHUB_WORKSPACE}/external/llvm-external-projects/torch-mlir-dialects" \
-DLLVM_TARGETS_TO_BUILD=AArch64 \
-DLLVM_USE_HOST_TOOLS=ON \
-DLLVM_ENABLE_ZSTD=OFF \
-DMLIR_ENABLE_BINDINGS_PYTHON=ON \
-DTORCH_MLIR_ENABLE_LTC=OFF \
-DTORCH_MLIR_ENABLE_MHLO=OFF \
-DTORCH_MLIR_USE_INSTALLED_PYTORCH="${{ matrix.torch-binary }}" \
-DMACOSX_DEPLOYMENT_TARGET=12.0 \
-DPython3_EXECUTABLE="$(which python)" \
$GITHUB_WORKSPACE/externals/llvm-project/llvm
- name: Build torch-mlir (cross-compile)
run: |
cmake --build build_arm64

1
build_tools/build_libtorch.sh
View File

@ -131,7 +131,6 @@ install_pytorch() {
unpack_pytorch() { unpack_pytorch() {
PYTHON_SITE=`${PYTHON_BIN} -c 'import sysconfig; print(sysconfig.get_paths()["purelib"])'` PYTHON_SITE=`${PYTHON_BIN} -c 'import sysconfig; print(sysconfig.get_paths()["purelib"])'`
pip uninstall torch
echo "wheel unpacking Pytorch..into ${PYTHON_SITE}" echo "wheel unpacking Pytorch..into ${PYTHON_SITE}"
wheel unpack -d "$WHEELHOUSE"/unpack_tmp "$WHEELHOUSE"/*.whl wheel unpack -d "$WHEELHOUSE"/unpack_tmp "$WHEELHOUSE"/*.whl
mv "$WHEELHOUSE"/unpack_tmp/* "$PYTHON_SITE"/ mv "$WHEELHOUSE"/unpack_tmp/* "$PYTHON_SITE"/